AI Architect

Enterprises are past the point of asking whether AI belongs in their systems. The question now is which AI capabilities should exist, how much autonomy each one earns, how they integrate with systems of record and real business processes, and how anyone will know they are working. Answering those questions well, and writing the answers down so that engineering teams and coding agents can build against them, is the AI Architect's job.
You will design the target architecture of AI-native applications, Enterprise AI and Applied AI solutions for large clients across industries, from assistants and RAG systems to agents, agentic workflows and autonomous components. You will decide what the system should do and, just as importantly, what it should not. You will define the evaluations, guardrails, and constraints that shape what gets built, whether the builder is an engineer or an AI coding agent.
This is an architecture role. The output is decisions, structure, and the reasoning behind them
- AI, not code volume. You will still prototype when a decision needs evidence, but most of your time goes to framing problems, resolving conflicting requirements, making trade-offs explicit, and getting the right people to own them.
Req# 1090267686The recruiting efforts for this position are intended to fill an existing vacancy for a new position.
ResponsibilitiesDesign end-to-end architectures for AI-native solutions: orchestration, retrieval, tool and system integration, agent autonomy and guardrails, identity and authorization, evaluation, observability, and cost modelFit each solution exactly to the business scope AI no more, no less AI and be able to explain what you deliberately left out and whyTurn incoherent or contradictory requirements into explicit business decisions: surface them, reframe them in business terms, price the options, and get a named owner to decideMake and document architectural trade-offs (build vs. buy, model selection, retrieval strategy, autonomy level, cost vs. quality) in a form that stakeholders can challenge and future teams can trustDefine the evaluation strategy for each system: how we will know that an assistant or agent fits the business scope, not just that it produces fluent outputAuthor the constraints, guidelines, and architectural fitness functions that engineering teams and coding agents build within, and revise them as the organization learns what generation gets wrongDecide where humans must stay in the loop AI regulated decisions, data of record, security boundaries, anything with legal or contractual consequenceWork within EPAM delivery teams and in presales: shape proposals, present options and costs to client executives, and review designs across teamsBuild proofs of concept and reference implementations when, and only when, a decision needs evidence or a team needs an example of \"good\"Contribute to EPAM's AI architecture practice: reference architectures, reusable patterns, and mentoring architects who are transitioning into AI-native designRequirementsYou have architected AI-native systems in production AI RAG, assistants, agents or agentic workflows with real users or real business processes behind them AI and you held architect-level responsibility for the system around the modelYou are a proven solution or software architect with a strong production track record who has started building with AI on your own initiative in the last year or so AI a side project, a proof of concept, an internal tool, published writing AI and wants to make AI-native design the center of your workSolid experience as a solution, software, or enterprise architect with delivery accountability: systems you designed that shipped, ran, and were maintained over timeDepth in distributed systems, integration, cloud, and data-flow architecture at enterprise scaleDemonstrated ability to work with ambiguous, conflicting requirements and to make trade-offs explicit rather than resolving them silentlyExperience presenting options, costs, and risks to non-technical decision makers and getting decisions madeA habit of writing things down: architecture decision records, design documents, reference architectures that other teams built fromRecent hands-on coding ability, enough to prototype and to keep your constraints honestUnderstanding of how AI-assisted development changes the architect's role: what to specify, what to evaluate, and where to keep human judgmentNice to haveExperience in regulated industries AI financial services, healthcare and life sciences, insuranceAgent frameworks, tool and function calling, MCP, orchestration patterns, LLM observabilityIdentity and authorization for agentic systems (OAuth 2.1, SPIFFE, AuthZEN, DPoP)Ownership of cost and token economics for a production AI systemConsulting or pre-sales experiencePublished writing, talks, or open-source contributions on AI architectureWe offerExtended Healthcare with Prescription Drugs, Dental and Vision, and Healthcare Spending Account (Compan